CoolFace
Modelpublic

thoughtworks/Nemotron-3.5-30B-A3B-Antislop-FTPO-LoRA

sourceHugging Faceotherupdated 2mo agoView on Hugging Face
2likes23downloads
Model Card

Nemotron-3.5-30B-A3B-Antislop-FTPO, LoRA adapter

The 842 MB LoRA adapter produced by running Antislop and FTPO against NVIDIA's Nemotron 3.5 30B-A3B. This repo holds the training delta on its own.

To run the model, use the merged checkpoint instead: thoughtworks/Nemotron-3.5-30B-A3B-Antislop-FTPO. That repo carries the full model card and benchmark tables. This one covers what is specific to the adapter.

Configuration

PEFT typeLoRA
Target moduleslm_head only
Rank (r)256
Alpha256
Dropout0.05
Trainable paramsabout 842 MB in BF16

Targeting lm_head alone is deliberate. FTPO adjusts final-token logits, so the output projection is where the preference lives, and constraining training to it keeps the rest of the model's capabilities intact. The same constraint caps achievable suppression. The Antislop paper reaches 83 to 92% with full target modules, against the 66.41% measured here, which matches its lm_head-only precedent on Llama-3.3-70B.

Usage

python
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base_id = "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16"
model = AutoModelForCausalLM.from_pretrained(
    base_id, torch_dtype="bfloat16", device_map="auto", trust_remote_code=True
)
model = PeftModel.from_pretrained(model, "thoughtworks/Nemotron-3.5-30B-A3B-Antislop-FTPO-LoRA")
model = model.merge_and_unload()

tok = AutoTokenizer.from_pretrained("thoughtworks/Nemotron-3.5-30B-A3B-Antislop-FTPO-LoRA")

The tokenizer, chat template, and special-token map bundled here are byte-identical to the base model's, included so the adapter is self-sufficient.

Headline result

Measured with the merged checkpoint on 400 held-out prompts, Antislop sampler off:

MetricBaselineFTPO
Banlist suppression (prose only)0%66.41%
Writing quality (0 to 100, n=150 paired)54.3053.34 (n.s.)
MMLU (600 q)0.83830.8433
GSM8K (250 q)0.92400.9360

Full tables and agentic benchmarks are in the merged model card.

What the adapter suppresses

The 4,267-pattern banlist is measured from this model's own output against a human baseline, not hand-written. Raw counts across the 400 held-out prompts, sampler off:

Banned patternBaselineFTPO
said, his voice dropping180
panic, cold and sharp150
heart hammered against my ribs121
dust motes dancing112
sharp, metallic tang50
smelled of ozone and old paper50
Elias (as protagonist name)24444
Kael20447

More examples, including assistant register bleeding into fiction, are on the merged model card.

Not included

The frozen FTPO reference adapter used during training (ref/, 803 MB) is not published here. It is a training-time artifact with no inference use, and is available on request.

License

OpenMDW-1.1, matching NVIDIA's public Nemotron 3.5 Lightning releases. The Antislop framework is MIT-licensed.

Citation

The method was published at ICLR 2026:

bibtex
@inproceedings{paech2026antislop,
  title     = {Antislop: A Comprehensive Framework for Identifying and Eliminating
               Repetitive Patterns in Language Models},
  author    = {Paech, Samuel and Roush, Allen and Goldfeder, Judah and Shwartz-Ziv, Ravid},
  booktitle = {The Fourteenth International Conference on Learning Representations},
  year      = {2026},
  url       = {https://openreview.net/forum?id=gLcyM1khyp},
  eprint    = {2510.15061},
  archivePrefix = {arXiv},
  primaryClass = {cs.CL}
}